Primer

Designing one view of every student's weekly progress

Before designing a consolidated goals interface, a student's progress lived inside each learning app, and every app has its own way of keeping tally. Khan counts points, IXL counts skills, Newsela counts quiz scores, independent reading counts minutes. To see where a student's progress was for that day or week, a teacher had to open each app individually. Mapped across an entire classroom meant it was nearly impossible to keep track of every student.

The first version of the design consolidated all the apps into one view: every student's daily goals, color-coded, green meant goal met, red meant goal missed, and grey meant no data. It solved the problem we set out to solve. Teachers could finally see the class at a glance.

But once teachers used it, the problem got deeper. A red icon tells you a student didn't meet their goal. It doesn't tell you why, or what to do about it. Green says the goal is met, but not by how much. The binary, green versus red, answered the first question: Who met their goals today? It didn't answer a more complicated question. What did they do today? The instinct to answer it by adding more to the dashboard would have buried the glance that made it useful.

The consolidated goals interface spanned four versions across a year, each one a different answer to the same widening question: How much can a single view carry before it stops being glanceable? The UX evolved as we designed, released and learned from teachers. The versions were:

1. A flat status grid
2. A proactive interpretation layer
3. Two layers of progressive disclosure
4. A week-level view with collapsed daily views and semantic icons

Along the way as other parts of the product were evolving, pieces of information kept migrating to wherever they were most useful, some onto a shared class schedule, some onto a large-screen classroom display. I designed each version, watched it as it was released to real classrooms, and used teacher feedback to evolve the next version.

Once the data being displayed was set, we learned from teachers. Using progressive disclosure, the teacher can hover on any UI element to better understand completion. And on click, a teacher can open any student's full app goals to see domain-level mastery and a lesson-by-lesson history, scores and time included, without ever opening the source apps.

Using an LLM to help design a project based learning class

Most goals came from outside apps. The exception was Pursuits, Primer's term for project-based learning. Teachers authored these classes themselves, and the goal was a daily habits-of-work score that lived in the same goal system as Khan and IXL. Before the platform supported a self-serve Pursuit builder, teachers wrote the courses in docs and submitted them to a person for review, which wouldn't scale as campuses were added.

I designed an LLM class-builder where teachers create a course and get feedback in the same place. A teacher works through foundation planning and instructional design, then submits for review. On submit, an LLM checks whether the academic focus, the learning goal, and the demonstration of learning are in sync, and returns structured feedback on each: aligned or not, with a note on what to revisit. The teacher revises and resubmits.

The impactful UX decision involved how the LLM's feedback behaved. I designed it as a soft gate. It never blocks a teacher from working, but a course only earns the 'ready to teach' label once it passes the LLM review, and a teacher who moves ahead without passing routes to a manual review queue instead.

That inverted the workload. A person used to read every course; now the LLM carries that volume and human review becomes the exception, which is what let course authoring scale across 18 campuses. Once a course passes, the builder unlocks more detailed authoring. I designed the flow and the loop, not the model itself.

A classroom made for teaching, not meetings

Most learning happened in the physical classroom, but some students logged into a digital one to work with peers at their learning level. When I joined, that ran on Zoom, which had two problems. It cost more at network scale, and it was made for meetings, not teaching. Zoom had no concept of a class roster, a lesson, or the small tools a teacher reaches for mid-instruction.

I designed Prism, a virtual classroom shaped around how a class actually runs: a fixed-seat roster, small-group instruction with custom class layouts, class and individual student chats, screen sharing, and controls to manage class noise. The tool list changed as we put it in front of teachers, and we kept iterating. Prism still runs today.

Keeping parents in the loop

The last piece of a teacher's day reached outside the classroom. Pulse gave parents a regular window in. I designed it as a per-class feed: teachers post photos and short updates, and parents see them on the native app. Teachers can also post field trip reminders, holiday schedules, and testing dates. Parents respond with emojis or comments, giving the feed a two-way dialogue. A regular view into what their kids were doing helped build trust with parents.

What changed for teachers

Everything a teacher needed lived in the platform. Each of these replaced a tool rather than adding one, except the class-builder, which added a capability the platform never had. The teacher experience also included attendance, grading, literacy class tools, and reporting. One centralized place to run a classroom, not a collection of stitched together tools.

What do I need to do right now?

Self-direction starts with knowing where you are in the day. I designed the schedule block by block, from opening circle to breaks, with the current block always highlighted as the day moves through it, and its apps surfaced right there. A student can go back to finish work they missed, but not jump ahead. The day only offers what's happening now, and that's what lets a student follow it without asking.

Every goal in one place

Knowing what to do now is half of it. A student also needs to see what they're working toward, and that lived across five apps, each with its own scoreboard: Khan in points, IXL in skills, Reflex in minutes, Newsela in quiz scores, reading in minutes. I pulled all five into one view, same card, same daily and weekly bars, so a student sees what to hit today and how the week is tracking without translating between systems.

A light reward sits on top: a coin for each daily goal met, spent at a midday reward shop on pencils, erasers, custom cursors and fonts. Keeping the shop closed the rest of the day means the coins reinforce the work without competing with it.

Designed for the screen the experience runs on

Teachers ran the classroom on a MacBook Air. Students ran it on Chromebooks, shorter screens with poorer color, in brightly lit rooms. I tested on a Chromebook for sizing and contrast under that lighting, and rather than shrink the teacher layout, I designed the student view for the screen it actually lived on. For Prism, the digital classroom, that meant keeping video, chat, and the current activity usable in far less vertical space.

What changed for students

Students knew where to be and when, what was expected each day and week, and they earned coins along the way. That gave them the autonomy to answer two questions on their own: what should I be doing right now, and what am I working toward. Which freed teachers to help the students who needed it.